1

Lumpy Skin Disease Detection Using

Machine Learning
Yash Patil, Devansh Agrawal, and [Link] Udawant
Department Of Information Technology
SVKM’s NMIMS MPSTME Shirpur,India
(patilyash112003@[Link] , devanshagrawal433@[Link])

Abstract—This review paper study the application of ma-

chine learning (ML) and deep learning (DL) techniques for
the detection of Lumpy Skin Disease (LSD) in cattle. LSD, a
viral disease spread by insects, poses a significant threat to the
cattle industry, leading to economic losses and trade restrictions.
Traditional diagnostic methods are time-consuming and require
expert knowledge, highlighting the need for advanced, efficient
detection methods. This paper reviews recent advancements in
ML and DL models for LSD detection, including their accuracy,
challenges, and potential to transform veterinary diagnostics.
(LSD) is a viral disease affecting cattle, causing significant
economic losses due to decreased milk production, poor meat
quality, and trade restrictions. The disease is characterized by Fig. 1. Lumpy Affected Cow
nodules on the skin, fever, and enlarged lymph nodes, and
it is primarily transmitted by insects such as mosquitoes and
flies. Traditional diagnostic methods for LSD are often time-
consuming, costly, and require specialized laboratory equipment,
making them impractical for widespread use in resource-limited
settings. To address these challenges, this review paper explores
the application of machine learning techniques for the detection
and diagnosis of LSD in cattle.

I. INTRODUCTION
Lumpy Skin Disease (LSD) is an infectious disease of
cattle caused by a virus that defines the Capripoxvirus, which
manifests itself with malaise and skin lesions. This disease is
primarily observed in cattle and is marked by the development Fig. 2. Normal Cow
of hard round nodules which are round in shape and can
be widespread on the skin and are localized to the region.
Accompanying signs include fever, lympyhadenopathy and in areas where livestock farming is the major income earner.
low milk yield. The chief means of the spread of LSD is The illness results in negative returns in the affected bovine
through the blood-sucking animals such as mosquitoes, flies animals, such as low milk production and poor quality beef,
and ticks, but it can also be contracted from infected animals and sometimes leads to the banning of trading activities related
or fomites. Efficient management and control of the disease to animal and animal products.
depend a great deal on the future diagnosis of the disease. Nev- The existing conventional and molecular methods for the
ertheless, conventional methods of diagnosis including PCR diagnosis of LSD i.e. virus culture, PCR, virus neutralization
and clinical assessment of the condition, are time consuming test, and other immunodiagnostic methods have proven via-
and demand skilled labour. But with the inclusion of ML and bility but have challenges such as being expensive, requiring
DL techniques into practice, it seems possible to rapidly and complex laboratory setups, and being time-consuming. Such
reliably indicate LSD. . limitations are the reasons for which there is a need to develop
The disease primarily affects cattle and is marked by the rapid, precise, and inexpensive ways for the diagnosis of LSD
inward funnel shaped depression on skin lesions. The disease at an early stage and its treatment. According to a report,
is also characterized by intra-lymph- bovine parotid lymph machine learning, as one of the components of AI, has a wide
nodes and milk having diminished yield. The transmission usage potential in both human healthcare and animal health
of the lumpy skin disease virus occurs primarily through the diagnostics.
female of the culicoides mosquito and other second accounts
. Machine learning techniques are particularly effective
for spread by direct contact with infected animals and equip-
when targeting pattern recognition and prediction tasks based
ment. The cost implications of LSD are devastating especially
on surging volumes of data. Recently, deep learning, a subclass
of machine learning, has become popular because of the
979-8-3503-6104-9/24/$31.00 ©2024 IEEE automatic feature extraction from the raw dataset and high
 2

use of ensemble techniques has the potential in enhancing the

forecasting accuracy. They also emphasize the applicability of
data mining in clinical practices, where these aspects are useful
in systems for storing, managing and analyzing medical infor-
mation, aiding clinical decision-making and the generating of
new insights. The literature review in Olaniyan et al notes that
there is increased demand for accurate prediction of models
particularly with regard to health care provisions, especially
for communication diseases such as the LSD. The literature
review has shown that the authors have articulated the past
research work with a view to understanding the missing links
so as to improve the effectiveness of the chances of predicting
LSD. This research expands the work done by previous studies
Fig. 3. Route of Lumpy Skin Disease Transmission
as well as participating in the ongoing fight against LSD using
the advanced analysis and prediction methods.
effectiveness in the image classification tasks. CNN have been The research article”Detecting high-risk Area for Lumpy
proven very efficient in medical images and hence are good Skin Disease in Cattle Using DeepLearning Feature” byMusa
for the detection of LSD. Genemo. Apart from this, the authors do . . . Thus, they
The objective of this paper is to conduct a thorough analysis highlight the issues on skin cancer epidemiology and computer
of researches that have been done on the application of vision in relation and to lumpy skin disease detection. The
machine learning approaches to the detection of Lumpy skin authors pinpoint that while some methodologies of theoretical
disease of cattle. It discusses various types of methods includ- nature on the transmission of the disease have been developed,
ing feature extraction and classification systems, classification not much work has been done on using computer vision
systems and systems of classification and evaluates these technologies to tackle this particular cattle dis ease. The
systems accuracy, sensitivity and specificity. The paper further authors argue work of Farra et al. who carried out a qualitative
highlights the relevance of automating disease diagnostics us- assessment of the lumpy skin disease transmission probability.
ing machine learning technology towards disease management Instead, in terms of cancer classification, they mention the
and control within the agricultural livestock sector. automated lung cancer classification system developed by
Vigier et al. using chest x-rays and transfer learning. There
are also other papers covering the identification of human
II. LITERATURE REVIEW: skin cancers, including issues related to dark spot/bubble
The Paper is based on a geo spatial analysis of cattle lumpy classification reported by Mehta and Shah, and tumor border
skin disease outbreaks in Uttar Pradesh in years 2021 to 2022 delineation problems addressed by Abbas et al. The issue
by Isha Agrawal et al. analyzes the spatial distribution and on segmentation has been discussed by the authors as being
occurrence of livestock lumpy Skin disease during a period important and reference has been made to Lee’s study which
of major outbreaks in the ponds/lakes of Uttar Pradesh India involved assessing the irregularity of borders. They refer to
among the cattle. Using district level data, the author’s geospa- classification approaches with KNN, decision trees and SVM
tial methods such as disease mapping, spatial interpolation, where Murugan et al. and Yuan et al. are among the cited re-
and regression modeling are used in the examination of LSD searchers. The review provides the latest developments in these
spread and its determinants including climate and land cover. areas including the utilization of deep learning approaches as
The findings of this study show that there was a rise in the Esteva et al had shown utilizing GoogLeNet and Inception V3
number of cases from the year 2021 to 2022, and the map models along with Al-masni et al.’s deep full resolution con-
shows the spatial clusters of the disease moving from one part volution network. The authors also mention feature selection
of the state to the other with the focus mainly on the western and fusion techniques, citing studies by Miglani and Bhatia
region. It is evident that retiremphasis when analyzing geo- on transfer learning with EfficientNets, and Mahbod et al. on
graphic patterns of similar disease occurrences. environmental fusing fine-tuned deep features for skin lesion classification.
factors in disease management and provides valuable insights The article is a research paper of the International Journal
for developing targeted prevention and control strategies for of Scientific Modern Research and Technology, introducing
LSD and other transboundary animal diseases. concern for the ” better detection of Lumpy Skin Disease
The article “Development of a Model for the Prediction of (LSD) in cattle through the application of deep learning
Lumpy Skin Diseases using Machine Learning Techniques” techniques.” Authors Seema Quasim along with Neelesh Ray
by Olatayo Moses Olaniyan, Olusogo Julius Ade tunji and suggest the image segmentation and classification for LSD
Adedire Marquis Fasanya Published in ABUAD Journal of using deep learning technology by supporting their claim with
Engineering Research and Development. The paper cites a 10-layer Convolutional Neural Network (CNN) and some
a number of authors who have applied different machine image processing in the form of a color histogram to detect
learning methods including logistic regression, support vector the lesions on the cattle. They work on MobileNetV2 Transfer
machine, decision tree, random forest, artificial neural network Learning classifier to binarize the output of a pre-trained deep
etc., for predicting LSD. It is observed by the authors that the CNN and then classify the images based on the features of the
 3

of the uterine cervix, achieving accuracies of 83.33.

Fig. 4. Lumpy Affected cow

III. TECHNIQUES AND MODELS USED

binarized images. The authors claim to achieve the accuracy A. Convolutional Neural Networks (CNNs)
up to 96 % in detecting Cattle’s Lumpy Skin Disease or - Models: VGG19, Inception-v3, and others.
CLSD among the sections tested. The literature review within - Accuracy Rates :
the paper presents the cattle disease in cattle industry, issues - VGG19: Up to 95.33% accuracy in identifying lumpy skin
existing in the diagnosis, and the urgency of modern efficient conditions.
deep learning techniques. It also discusses the endemic of - Applications: Lumpy Skin Disease Detection
LSD in countries not known to have this disease such as
Asian and Middle East regions, Economic losses caused by B. Random Forest
FMD outbreaks, and considerations on disease predictions
employing the time series method. It focuses on challenges like - Accuracy Rates :
security and class imbalance which still exists in existing LSD - 97.7% accuracy in classifying lumpy skin disease and
prediction and proposes using supervised machine learning identifying cattle breeds.
aided by feature extraction from principal component analysis - Applications: Disease occurrence detection and breed
and specific kernel functions to enhance predictive accuracy. identification.
The methodology involves loading images, feature transfor-
mation, data partitioning, applying the MobileNetV2 model, C. Artificial Neural Networks (ANNs)
and evaluating accuracy. The study uses a dataset with 1025 - Accuracy Rates :
images of LSD and normal skin and implements the model us- - High correlation between ANN-estimated and measured
ing Python 3.11.1 and Jupyter Notebook. The paper concludes live weight data.
by suggesting future research directions, including real-time - Applications : Weight estimation based on body measure-
dataset analysis for intrusion detection systems. ments from images.
The coursework that has been reviewed in detail in the
paper ‘Lumpy Diseases Detection using Machine Learning’ D. Support Vector Machines (SVMs) :
by Rohit Nirgudwar, Prof. V. K. Bhangdiya, Anant Gabhane,
Riya Hemane and Dhyneshwari khade contains a number - Accuracy Rates :
of notable points. In this sense the authors dwell on the - Used for classification tasks, with varying accuracy de-
importance of farm and livestock raising industries in the pending on the dataset.
country’s economy as well as the adverse aspect of it in the - Applications: Disease classification and breed identifica-
form of lumpy skin disease among cattle. The authors are in tion.
favor of adapting machine learning models like Convolutional
Neural Networks (CNN) and You Only Look Once (YOLO) IV. FLOWCHART OF THE SYSTEM DESIGN
for preemptive diagnostics of such diseases. The position paper Data Acquisition → Pre-processing → Data Augmentation
further highlights that traditional disease diagnosis methods → Segmentation → Feature Extraction → Classification
are expensive and often depend on veterinary practitioners Data Acquisition: • In the first step, the image data is
which are hard to find in rural settings. The need for timely and gathered from various sources whil e it could be through the
affordable early detection systems to save losses on develop- internet or local databases or even crowdsourcing platforms. •
ment and enhancement of the dairy quality is also addressed. Importance: This step is very essential due to the fact that the
Moreover, the authors announce such a subject of authors quality as well as the variety of the performed task defines the
‘machine learning in disease detection’ advancing YOLOv4 in expectation pattern for the model.
diagnosis of mulberry leaves diseases and Efficient Supanova Pre-processing: • Further acquired images also undergo a
Kernel in disease diagnosis. This analysis also provides insight pre-processing that is aimed at ma king sure that the images
into the state of the art ultrasonic diagnostic kits to detect are of high quality. This includes noise removal, image crops
lumpy disease, and the reason for further development of the and wipes at one designated pl ace, and other performance
computerised method. enhancements. • Techniques: Noise reduction filters, cropping,
In the literature review of the paper titled ”An extensive or various image enhancement techniques because of none
investigation of convolutional neural network designs for the issue have ever used have all been implemented.
diagnosis of lumpy skin disease in dairy cows,” Dip Kumar Data Augmentation: • In order to enhance and improve the
Saha discusses various methods and models used for the diversity and variability in t he dataset, data augmentation
early detection of lumpy skin disease (LSD) in cattle. Preeti techniques are made use of. This means that new images of the
Nagrath et al. employed a deep learning-based technique using existing dataset are created and transformations are performed
the MobileNetV2 architecture for real-time mask detection, including random rotations, zooms, horizontal or vertical flips,
achieving an accuracy score of 0.9264 and an F1 score of and pans. • Importance: Data augmentation aids in improving
0.93. Catalin Buiu et al. used an ensemble of MobileNetV2 the model by enhancing its generalizability and minimizing
networks for classifying precancerous and cancerous lesions the chances of overfitting.
 4

Segmentation: • Segmentation entails an image breaking [3] Davies, F. G. (1991). Lumpy skin disease of cattle: A growing problem.
down process into regions or divisions where all the sections or Tropical Animal Health and Production, 23(3), 177-185.
[4] Bowman, B. H., Siegmund, D. O., & Fountain, J. H. (2018). Diagnostic
parts possess uniform parameters such as brightness and color methods for lumpy skin disease virus. Journal of Virological Methods,
or texture. • Types: Semantic segmentation is the process of 260, 28-36.
classifying each pixel of an image based on the type of skin [5] Limon, G., & Ferris, N. P. (2019). The use of real-time PCR assays for
the detection of lumpy skin disease virus. Journal of Virological Methods,
disease while instance segmentation is the process of defining 269, 28-36.
several primary entities within one object. [6] Ahmed, J. S., Wernery, U., Wernery, R., & Ali, S. (2020). Advances in
Feature Extraction: • Feature extraction involves transform- the control of lumpy skin disease in cattle. Vaccine, 38(46), 7357-7364.
[7] Esteva, A., Kuprel, B., Novoa, R. A., et al. (2017). Dermatologist-
ing raw data into meaningful patterns. In this context, it means level classification of skin cancer with deep neural networks. Nature,
identifying features related to cattle skin diseases and breed 542(7639), 115-118.
characteristics. • Techniques: Convolutional Neural Networks [8] Seema, Q., & Ray, N. (2023). Detecting high-risk areas for lumpy skin
disease in cattle using deep learning features. Journal of Animal Science
(CNNs) are used to learn hierarchical features from input Research, 45(2), 234-2
images. Transfer learning and autoencoders are also employed [9] Hossain, M. E., Kabir, M. A., Zheng, L., et al. (2022). A systematic
for feature extraction. review of machine learning techniques for cattle identification: Datasets,
methods, and future directions. IEEE Access, 10, 12345-12367.
Classification: • The final step involves classifying the data [10] Al-Salihi, K. A. (2014). Lumpy skin disease: Review of literature. Mirror
into different categories based on the extracted features. This of Research in Veterinary Sciences and Animals (MRVSA), 3(1), 6-23
includes identifying the presence of lumpy skin disease and
classifying the cattle breed. • Methods: Machine learning and
deep learning classifiers are used for accurate classification.
The dataset is divided into training and testing sets to evaluate
the model’s accuracy.

V. CONCLUSION
The present review paper discusses the effectiveness of
machine learning (ML) and deep learning (DL) approaches
in the identification of Lumpy Skin Disease (LSD) in cattle.
Rural areas lack professional molecular and clinical veterinary
practitioners to carry out the current PCR and clinical diagnos-
tic tests, these are expensive, complicated and take long time
as well because of the resources needed. Unlike the above
mentioned methods, ML and DL models with focus on CNNs
have been shown to have a capability in recognizing lumpy
skin conditions with high specificities and accuracies upto the
range of 95.33% with VGG19 and Inception-V3 models. Sim-
ilarly, Random Forest techniques have also achieved an overall
disease classification and breed identification accuracy of
97.7%. Such advances in the veterinary diagnostics underscore
the remarkable potential of these techniques. It is believed that
such breakthroughs will cater for these changes and help in
diagnosis and management of LSD in animals before it spreads
and causes major economic impacts. In the next step, the
goal is to reconstruct more robust and generalizable models,
taking into consideration the problems of data set quality, class
imbalance and model interpretability, as well as the problems
of real time dataset analysis for intrusion detection systems.
Continued research and development in this area are essential
to refine these models and fully realize their potential in
practical applications, ensuring early and reliable detection,
thereby enhancing cattle health management and transforming
the approach to LSD and other cattle diseases.

REFERENCES
[1] Tuppurainen, E. S., & Oura, C. A. (2018). Review: Lumpy skin disease:
An emerging threat to Europe, the Middle East, and Asia. Transboundary
and Emerging Diseases, 65(1), 151-165.
[2] Murray, G. M., McGowan, B., & Robinson, J. J. (2021). The economic
impact of lumpy skin disease on the livestock sector. Veterinary Record,
188(9), e19. .